DocsAgentsAgent Playground

Agent Playground

Test and refine your agent's responses in a sandbox environment.

Debugging Responses

The Playground includes a "Show Debug Info" toggle. Enabling this reveals:

  1. Confidence Score: How sure the AI is about its answer (0.0 to 1.0).
  2. Retrieval Sources: The exact snippets of text the AI used to generate the answer.
    • Tip: If the AI hallucinates, check if the source text is misleading or outdated.

Testing Best Practices

  • Ask Edge Cases: Try questions your users rarely ask to ensure safety.
  • Test Formatting: Check if lists and links render correctly.
  • Verify Tone: Ensure the custom instruction (e.g., "Be professional") is being followed.

Was this page helpful?

Edit this page on GitHub